Winter air pollution studies mostly ignore the viruses circulating at the same time

Air pollution peaks in winter in much of the Northern Hemisphere. So do respiratory viruses. Time-series studies of pollution and emergency visits routinely adjust for season, weather and long-term trend — but far fewer account for how much virus was actually circulating on a given week. Two papers published this month put that variable into the model, and they do not agree about what it does.

Amplification at peak influenza

The first, published in Scientific Reports on 19 February, analysed daily emergency department visits for all respiratory illnesses, pneumonia, asthma and chronic obstructive pulmonary disease across six US metropolitan areas from 2005 to 2017 [s1]. Weekly influenza activity was measured using laboratory-confirmed influenza hospitalisation rates from a hospital surveillance system, and quasi-Poisson models estimated associations between PM2.5 and NO2 and ED visits during the influenza season [s1]. Effect modification was tested through quartile-level and spline-based interaction terms [s1].

PM2.5 was consistently associated with all respiratory outcomes; NO2 associations were more variable across locations [s1]. Adjusting for influenza activity slightly attenuated the pollution associations — the confounding contribution was small [s1].

The modification was larger. In Atlanta, the relative risk of respiratory ED visits per interquartile-range increase in PM2.5 was 1.035 (95% CI 1.020–1.051) during peak influenza periods, defined as the fourth quartile of activity, against 0.998 (95% CI 0.975–1.021) during low influenza periods [s1]. In San Francisco the corresponding figures were 1.013 (95% CI 1.006–1.024) and 0.999 (95% CI 0.993–1.006) [s1]. The highest NO2 associations also occurred during peak influenza periods in pooled analyses of pneumonia and COPD visits [s1].

The authors describe the modification as positive but complex and non-linear, varying by pollutant, outcome and location, with amplification during moderate or high influenza circulation [s1]. Their conclusion is a call for integrated air quality and infectious disease surveillance [s1].

Attenuation when rhinovirus is high

The second study, published in Environmental Health on 25 February, asks a narrower version of the same question in children and reaches a different answer [s2].

It analysed 12,603 emergency department visits for wheeze-associated disorders — acute asthma and virus-induced wheezing — among children aged 2 to 18 across eight hospitals in the Netherlands between 2016 and 2023 [s2]. Quasi-Poisson models estimated associations between same-day nitrogen dioxide, PM2.5, PM10 and ozone and daily visits, adjusted for seasonality, meteorology and pollen [s2]. Base models were then compared with models including weekly rhinovirus or respiratory syncytial virus positivity ratios as covariates or as interaction terms [s2].

In the base models, 3-day lag exposure to NO2 and PM2.5 was associated with an increase in visits: NO2 excess risk 2.9% (95% CI 0.4–5.6; p = 0.025) and PM2.5 excess risk 3.6% (95% CI 0.4–6.9; p = 0.026) [s2]. As in the US study, associations were similar after adjustment for viral activity — meaning viral circulation was not acting as a confounder [s2].

But the interaction models produced attenuated effect estimates for NO2, PM2.5 and PM10 during periods of relatively high rhinovirus activity [s2]. Ozone showed a negative association during low rhinovirus periods [s2]. The authors conclude that circulating rhinovirus modifies, rather than confounds, the association between short-term air pollution and paediatric wheeze visits, and that accounting for viral activity improves interpretation and may explain inconsistencies across studies [s2].

Why the two results are not necessarily contradictory

The studies differ in almost every dimension that would matter. Different countries and pollution regimes; different age ranges — all ages versus children aged 2 to 18; different outcomes — all respiratory illness, pneumonia and COPD versus wheeze-associated disorders specifically; different viruses — influenza versus rhinovirus and RSV; and different periods [s1][s2].

There is also a plausible mechanism for attenuation that does not require pollution to become less harmful. When a large share of paediatric wheeze presentations is being driven by rhinovirus, additional visits attributable to a marginal increase in particulate matter form a smaller proportion of a bigger total. That is a statistical dilution effect, not a protective one — and it is precisely the kind of ambiguity that a single point estimate from a model without virus data would hide.

What both studies establish, independently, is the narrower and more robust claim: viral circulation is an effect modifier rather than a confounder in these associations [s1][s2]. Studies that omit it are not producing biased estimates so much as producing averages across conditions in which the true effect differs.

Both are ecological time-series analyses using population-level exposure estimates and administrative visit records. Neither follows individuals, and neither can establish causation. The US study covers six cities; the Dutch study eight hospitals [s1][s2].

Which particles, from where

A third paper, published in Scientific Reports on 27 February, addresses a different gap: not when particles matter, but which ones [s3]. Using high-resolution dispersion models of particles from different local sources and address registries, the authors assigned annual individual residential concentrations to population-based cohorts in Gothenburg, Stockholm and Umeå from 1990 to 2011 [s3].

Among 68,679 participants there were 7,344 natural deaths, including 2,755 cardiovascular deaths [s3]. Exposure levels were moderate but generally above the WHO 2021 guidelines [s3]. Positive associations with natural mortality appeared for the last five years of exposure to road traffic exhaust particles (HR 1.02, 95% CI 1.00–1.04 per interquartile range; HR 1.10, 95% CI 1.00–1.22 per 1 µg/m³) and road wear particles (HR 1.02, 95% CI 1.00–1.04 per IQR; HR 1.02, 95% CI 1.00–1.03 per 1 µg/m³) — but not for particles from residential heating [s3].

Adjusting for road traffic noise, or for residential heating particles, did not substantially change the traffic results [s3]. For cardiovascular mortality specifically, associations with particles from both sources were positive but not statistically significant [s3]. The confidence intervals here touch 1.00 at their lower bound; these are weak positive signals, and the authors describe them as lending some support for further efforts to reduce traffic emissions [s3].

What to watch

Whether air quality and infectious disease surveillance systems are actually linked anywhere, which is the operational recommendation both time-series studies arrive at [s1][s2]. And whether source-apportioned exposure — traffic exhaust, road wear, residential heating treated separately — becomes standard in cohort studies, since a single PM2.5 figure averages over sources with apparently different mortality signals [s3].

Sources

1. [s1] Community-level influenza activity modifies the association between ambient air pollution and acute respiratory emergency visits in six U.S. Cities. Scientific Reports, published online 19 February 2026. https://doi.org/10.1038/s41598-026-39576-3
2. [s2] Seasonal viruses modify short-term air pollution effects on pediatric wheeze and asthma: a time-series study. Environmental Health, published online 25 February 2026. https://doi.org/10.1186/s12940-026-01274-y
3. [s3] Long-term exposure to particulate matter from road traffic and residential heating and mortality: a multi-cohort study in Sweden. Scientific Reports, published online 27 February 2026. https://doi.org/10.1038/s41598-026-37471-5